R Notebook

Feature selection strategies

Genes with high dispersion

cds_high_disp <- readRDS('output/monocle/180831/monocle_high_dispersion/10x-180831-monocle')
plot_cell_trajectory(cds_high_disp, color_by = 'timepoint')

DE genes between T1T2T3 and T4T5

cds_timecombined <- readRDS('output/monocle/180831/monocle_time-combined/10x-180831-monocle')
plot_cell_trajectory(cds_timecombined, color_by='timepoint')

DE genes from cluster resolution 1.5

cds_res1.5 <- readRDS('output/monocle/180831/monocle_res1.5/10x-180831-monocle')
plot_cell_trajectory(cds_res1.5, color_by='timepoint')

Dataset split into T1+T2+T3 and T4+T5, DE genes from clusters res0.5.

cds_split_res0.5 <- readRDS('output/monocle/180831/monocle_T1T2T3_T4T5_res0.5/10x-180831-noreg-monocle')
plot_cell_trajectory(cds_split_res0.5, color_by='timepoint')

Dataset split into T1+T2+T3 and T4+T5, DE genes from clusters res1.5 are used here. This trajectory was used for further analyses.

cds <- readRDS('output/monocle/180831/monocle_T1T2T3_T4T5_res1.5/10x-180831-monocle')
plot_cell_trajectory(cds, color_by='timepoint')

pdata <- pData(cds)

DE genes between brown and white.

cds_type <- readRDS('output/monocle/180831/monocle_type/10x-180831-monocle')
plot_cell_trajectory(cds_type, color_by='timepoint')

DE genes between top10% lowest PC2 score and top10% highest PC2 score.

cds_pc2 <- readRDS('output/monocle/180831/monocle_PC2/10x-180831-monocle')
plot_grid(
  plot_cell_trajectory(cds_pc2, color_by='timepoint'),
  plot_cell_trajectory(cds_pc2, color_by='State.old') + scale_color_manual(values=c("#f67770", "#964B00", "orange"), name = "State"),
  plot_cell_trajectory(cds_pc2, color_by='Pseudotime'),
  plot_cell_trajectory(cds_pc2, color_by='State'),
  ncol=2
)

plot_cell_trajectory(cds_pc2, color_by='timepoint') + geom_point(color='white', size=5) + geom_point(aes(colour=timepoint), alpha=0.01) 

plot_grid(
  plot_cell_trajectory(cds_pc2, color_by='State.old') + scale_color_manual(values=c("#f67770", "#964B00", "orange"), name = "State"),
  plot_cell_trajectory(cds_pc2, color_by='depot') + scale_color_manual(values=c("#845d1e", "#edd2a8", "#472809", 'orange'), name = "Depot"),
  ncol=2
)

Ratio’s white/brown and depots per branch.

#TODO rewrite so it outputs a table 
get_ratios <- function(col1, col2){
  states <- unique(seurobj@meta.data[,col1])
  values <- unique(seurobj@meta.data[,col2])
  for (state in states){
    n_state = length(which(seurobj@meta.data[col1] == state))
    print(paste('N cells', col1, state, ':', n_state))
    for (value in values){
        n_state_value <- length(which(seurobj@meta.data[col1] == state & seurobj@meta.data[col2] == value))
        perc_state_value <- n_state_value / n_state
        print(paste('Ratio', value, 'in state', state, ': ', round(perc_state_value, 2)))
    }
  }
}

states_pc2 <- pData(cds_pc2)['State']
names(states_pc2) <- 'State.PC2'
seurobj <- AddMetaData(seurobj, states_pc2)
get_ratios('State.PC2', 'depot')

get_ratios('State.PC2', 'State.old.labels')

get_ratios('State.old.labels', 'depot')

get_ratios('depot', 'State.old.labels')

plot_grid(
  TSNEPlot(seurobj, group.by='State.PC2', pt.size=0.1, colors.use = c('#1bb840', '#649efc', '#f67770')),
  TSNEPlot(seurobj, group.by='State.old.labels', pt.size=0.1, colors.use = c('#649efc', '#1bb840', '#f67770')),
  ncol=2
)